Learning From Noisy Anchors for One-Stage Object Detection

Li, Hengduo; Wu, Zuxuan; Zhu, Chen; Xiong, Caiming; Socher, Richard; Davis, Larry S.

doi:10.1109/cvpr42600.2020.01060

Cited by 93 publications

(35 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It happens quite often that people annotate inaccurate bboxes in images/videos as the ground truth, even with computer-assisted annotation tools. However, there is little work on the robustness of localization losses to noisy bboxes, even though a number of methods have been proposed for robust learning with noisy labels, anchors, and bboxes [27,26,12,10,15,37,38,25,19,20]. Here, we fill this gap by conducting a set of experiments to evaluate the robustness of different localization losses to noisy bboxes.…”

Section: Robustness To Noisy Bounding Boxesmentioning

confidence: 99%

Alpha-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression

Erfani

et al. 2021

Preprint

View full text Add to dashboard Cite

Bounding box (bbox) regression is a fundamental task in computer vision. So far, the most commonly used loss functions for bbox regression are the Intersection over Union (IoU) loss and its variants. In this paper, we generalize existing IoUbased losses to a new family of power IoU losses that have a power IoU term and an additional power regularization term with a single power parameter α. We call this new family of losses the α-IoU losses and analyze properties such as order preservingness and loss/gradient reweighting. Experiments on multiple object detection benchmarks and models demonstrate that α-IoU losses, 1) can surpass existing IoU-based losses by a noticeable performance margin; 2) offer detectors more flexibility in achieving different levels of bbox regression accuracy by modulating α; and 3) are more robust to small datasets and noisy bboxes.

show abstract

Section: Robustness To Noisy Bounding Boxesmentioning

confidence: 99%

Alpha-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression

Erfani

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Also based on the ATSS, Kim et al [30] proposes PAA with a new anchor assignment strategy, extending some ideas such as selecting positive samples based on the detection-specific likelihood [43], the statistics of anchor IoUs [28], or the cleanness score of anchors [44,45]. The anchor assignment may consider a flexible number of positive (or negative) not only based on IoU, but also how probable the assignment can argue by the model, in other words, how meaningful the algorithm finds the anchor about the target object (which may not be the highest IoU) to assign it as a positive sample.…”

Section: Object Detection Methodsmentioning

confidence: 99%

Active Fire Mapping on Brazilian Pantanal Based on Deep Learning and CBERS 04A Imagery

et al. 2022

View full text Add to dashboard Cite

Fire in Brazilian Pantanal represents a serious threat to biodiversity. The Brazilian National Institute of Spatial Research (INPE) has a program named Queimadas, which estimated from January 2020 to October 2020, a burned area in Pantanal of approximately 40,606 km2. This program also provides daily data of active fire (fires spots) from a methodology that uses MODIS (Aqua and Terra) sensor data as reference satellites, which presents limitations mainly when dealing with small active fires. Remote sensing researches on active fire dynamics have contributed to wildfire comprehension, despite generally applying low spatial resolution data. Convolutional Neural Networks (CNN) associated with high- and medium-resolution remote sensing data may provide a complementary strategy to small active fire detection. We propose an approach based on object detection methods to map active fire in the Pantanal. In this approach, a post-processing strategy is adopted based on Non-Max Suppression (NMS) to reduce the number of highly overlapped detections. Extensive experiments were conducted, generating 150 models, as five-folds were considered. We generate a public dataset with 775-RGB image patches from the Wide Field Imager (WFI) sensor onboard the China Brazil Earth Resources Satellite (CBERS) 4A. The patches resulted from 49 images acquired from May to August 2020 and present a spatial and temporal resolutions of 55 m and five days, respectively. The proposed approach uses a point (active fire) to generate squared bounding boxes. Our findings indicate that accurate results were achieved, even considering recent images from 2021, showing the generalization capability of our models to complement other researches and wildfire databases such as the current program Queimadas in detecting active fire in this complex environment. The approach may be extended and evaluated in other environmental conditions worldwide where active fire detection is still a required information in fire fighting and rescue initiatives.

show abstract

“…On MS-COCO test-dev, CentripetalNet not only beat all existing anchor-free detectors with 48.00% AP, but also achieves performance equivalent to the latest instance segmentation method with 40.21% MaskAP. Li et al [12] set a cleanliness score for each anchor. Reduce the impact of noisy sample on classification and positioning to adaptively adjust the importance of each anchor during training.…”

Section: Related Workmentioning

confidence: 99%

Deep Learning-Based Object Detection Improvement for Fine-Grained Birds

Yang

Song

2021

IEEE Access

View full text Add to dashboard Cite

When the object detection algorithm is applied to the bird protection project, there are many problems like large model parameters, high similarity between bird species and single sample scene. In order to further improve the detection accuracy and stability of the object detection model, a multi-object detection algorithm for fine-grained birds is proposed. Firstly, the algorithm introduces Depthwise separable convolution into the feature extraction layer of YOLOv3 algorithm. The convolution process is divided into two parts: deep convolution and point-by-point convolution. The separation between intra-channel convolution and inter-channel convolution is realized. On the basis of high detection accuracy, the number of algorithm model parameters and calculation amount are greatly reduced. Finally, Focal loss was added to the loss function to solve the serious imbalance of positive and negative samples. By reducing the weight of the large number of simple background classes, the algorithm was more focused on detecting foreground classes. The experimental results show that, in the bird data set, the average precision mean (mAP) index of this algorithm is 2.71% higher than YOLOv3 algorithm, the number of parameters is 79.88% lower than YOLOv3 basic model, and the number of frames per second (FPS) is 19.98% higher than YOLOv3 algorithm. This algorithm not only greatly reduces the number of model parameters and computation, but also improves the detection speed and mAP.

show abstract

Learning From Noisy Anchors for One-Stage Object Detection

Cited by 93 publications

References 28 publications

Alpha-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression

Alpha-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression

Active Fire Mapping on Brazilian Pantanal Based on Deep Learning and CBERS 04A Imagery

Deep Learning-Based Object Detection Improvement for Fine-Grained Birds

Contact Info

Product

Resources

About